Sorry about your issues.
In general, I agree with Wouter's recommendations. In particular:
- SEP detection
- Full field star detection with an inner annulus of 25% and an outer annulus of 80%
- For the moonlight v2 controller on my 4" refractor, I use a step size of 25
- 5% tolerance
- If you find the focus goes far away sometimes, you can constrain it with maximum travel. E.g. give it some freedom, but if, e.g. it should never travel 300 steps away, then use 300. For those without such issues, it's OK to leave that very large.
- I'm not as much a fan of multiple exposures if you're already doing full field, but, of course, feel free to experiment.
- Self serving for me, but if you're having trouble with the Polynomial algorithm, try the Linear algorithm: see Jasem's latest announcement about 3.4.0
knro.blogspot.com/2020/02/kstars-v340-is-released.html
Tolerance is the percentage of HFR worse than the estimated best focus HFR that the algorithm is willing to "settle for". Setting it too tight, and the algorithm may never find it because of noise in the measurements and backlash and so on. Also, I know my Moonlight has backlash, I'd be very surprised if yours didn't. It's just that the driver has no mechanism for repairing it. The Linear focus algorithm attempts to deal with it.
Someday we'll have a focus wizard to set all these parameters :/
Focus (and align) can fair for reasons that can't be fixed, e.g. a cloud comes by. I agree that the system should be able to recover from those. You bring up a valid point that the system should not try to re-align 196 times
I've thought that we should also have a maximum HFR parameter, e.g. the AF algorithm should fail if the best HFR was 5. Do you have a log for that 196x align run?
Hy